Back to EveryPatent.com
United States Patent |
6,201,175
|
Kikumoto
,   et al.
|
March 13, 2001
|
Waveform reproduction apparatus
Abstract
A method for generating a compressed or expanded waveform from original
waveform data is disclosed. A plurality of frequency band-divided waveform
data is generated from the original waveform data. Phase and amplitude
data having periodically changing amplitudes are extracted for each of the
plurality of frequency band-divided waveform data. The periodically
changing phase data is converted to periodically changing frequency data,
and a temporal change rate of the periodically changing amplitude and
frequency data is adjusted for each of the plurality of frequency
band-divided waveform data. The temporally adjusted amplitude and
frequency data for the plurality of frequency band-divided waveform data
is then combined to generate the compressed or expanded waveform.
Inventors:
|
Kikumoto; Tadao (Shizuoka, JP);
Hoshiai; Atsushi (Hannan, JP);
Kusakabe; Satoshi (Hamamatsu, JP)
|
Assignee:
|
Roland Corporation (Osaka, JP)
|
Appl. No.:
|
511009 |
Filed:
|
February 22, 2000 |
Foreign Application Priority Data
| Sep 08, 1999[JP] | 11-254569 |
Current U.S. Class: |
84/605; 708/203; 708/420 |
Intern'l Class: |
G06F 017/15; G10H 007/04 |
Field of Search: |
84/605
708/203,315,420,421
|
References Cited
U.S. Patent Documents
3946504 | Mar., 1976 | Nakano | 35/35.
|
4805217 | Feb., 1989 | Morihiro et al. | 381/35.
|
4876937 | Oct., 1989 | Suzuki.
| |
5315057 | May., 1994 | Land et al. | 84/601.
|
5347478 | Sep., 1994 | Suzuki et al. | 364/715.
|
5412152 | May., 1995 | Kageyama.
| |
5471009 | Nov., 1995 | Oba et al. | 84/645.
|
5499316 | Mar., 1996 | Sudoh et al. | 395/2.
|
5511000 | Apr., 1996 | Kaloi et al. | 364/514.
|
5611018 | Mar., 1997 | Tanaka et al. | 395/2.
|
5675709 | Oct., 1997 | Chiba | 395/2.
|
5713021 | Jan., 1998 | Kondo et al. | 395/614.
|
5717818 | Feb., 1998 | Nejime et al. | 395/2.
|
5734119 | Mar., 1998 | France et al. | 84/622.
|
5745650 | Apr., 1998 | Otsuka et al.
| |
5763800 | Jun., 1998 | Rossum et al. | 84/603.
|
5765129 | Jun., 1998 | Hyman et al. | 704/270.
|
5774863 | Jun., 1998 | Okano et al. | 704/278.
|
5781696 | Jul., 1998 | Oh et al. | 395/2.
|
5792971 | Aug., 1998 | Timis et al.
| |
5809454 | Sep., 1998 | Okada et al. | 704/214.
|
5847303 | Dec., 1998 | Matsumoto | 84/610.
|
5873059 | Feb., 1999 | Iijima et al. | 704/207.
|
Primary Examiner: Witkowski; Stanley J.
Attorney, Agent or Firm: Foley & Lardner
Claims
What is claimed is:
1. A waveform reproduction apparatus for generating a compressed or
expanded waveform from original waveform data, the original waveform data
being divided into a plurality of waveform components representing a
plurality of frequency bands, the apparatus comprising:
a storage means for storing temporally changing phase data and amplitude
data that represent phases and amplitudes of each waveform component;
a frequency data conversion means in communication with the storage means
for converting the phase data into frequency data;
a change rate adjustment means in communication with the storage means and
the frequency data conversion means for adjusting temporal change rates of
the frequency data and amplitude data; and
a waveform reproduction means in communication with the storage means and
the change rate adjustment means for generating output waveform data which
is a temporally compressed or expanded version of the original waveform
data by reproducing the stored frequency data and amplitude data in
accordance with the adjusted temporal change rates of the frequency data
and amplitude data.
2. A waveform reproduction apparatus as recited in claim 1, wherein the
change rate adjustment means adjusts the temporal change rate for the
amplitude data by duplicating the amplitude data for an amount equal to at
least one integer period of the amplitude data and adding or omitting the
duplicated amplitude data to the amplitude data.
3. A waveform reproduction apparatus as recited in claim 2, wherein the
change rate adjustment means adjusts the temporal change rate for the
frequency data by duplicating the phase or frequency data that corresponds
to the added or omitted amplitude data for an amount equal to at least one
integer period of the phase or frequency data, and adding or omitting the
duplicated phase or frequency data to the phase or frequency data.
4. A waveform reproduction apparatus for generating a compressed or
expanded waveform from original waveform data, the original waveform data
being divided into a plurality of waveform components representing a
plurality of frequency bands, the apparatus comprising:
a storage means for storing temporally changing frequency data and
amplitude data that represent frequencies and amplitudes of each waveform
component;
a change rate adjustment means in communication with the storage means for
adjusting temporal change rates of the frequency data and amplitude data;
and
a waveform reproduction means in communication with the storage means and
the change rate adjustment means for generating output waveform data which
is a temporally compressed or expanded version of the original waveform
data by reproducing the stored frequency data and amplitude data in
accordance with the adjusted temporal change rates of the frequency data
and amplitude data;
wherein the change rate adjustment means adjusts the temporal change rate
for the amplitude data by duplicating the amplitude data for an amount
equal to at least one integer period of the amplitude data, and adding or
omitting the duplicated amplitude data to the amplitude data.
5. A waveform reproduction apparatus as recited in claim 4, wherein the
change rate adjustment means adjusts the temporal change rate for the
frequency data by duplicating the frequency data that corresponds to the
added or omitted amplitude data, and adding or omitting the duplicated
frequency data to the frequency data.
6. A method for generating a compressed or expanded waveform from original
waveform data, the method comprising the steps of:
generating a plurality of frequency band-divided waveform data from the
original waveform data;
extracting phase and amplitude data having periodically changing amplitudes
for each of the plurality of frequency band-divided waveform data;
converting the periodically changing phase data to periodically changing
frequency data and adjusting a temporal change rate of the periodically
changing amplitude and frequency data for each of the plurality of
frequency band-divided waveform data;
reproducing the amplitude and frequency data in accordance with the
adjusted temporal change rates of the amplitude and frequency data for
each of the plurality of frequency band-divided waveform data; and
combining the reproduced amplitude and frequency data for each of the
plurality of frequency band-divided waveform data to generate the
compressed or expanded waveform.
7. A method as recited in claim 6, the step of generating a plurality of
frequency band-divided waveform data comprising dividing the original
waveform data into a plurality of frequency band-divided waveform data so
that each frequency band contains a center frequency and a plurality of
harmonics that are integer multiples of the center frequency.
8. A method as recited in claim 7, the steps of dividing the original
waveform data into a plurality of frequency band-divided waveform data and
extracting phase and amplitude data having periodically changing
amplitudes comprising the steps of:
for each of the plurality of frequency band-divided waveform data, and for
each of a plurality of N instantaneous values of the original waveform
data and a corresponding Nth data item of the center frequency received in
time sequence within each of the plurality of frequency band-divided
waveform data,
multiplying the Nth data item of the center frequency within a particular
frequency band-divided waveform data by the Nth instantaneous value of the
original waveform data to produce an intermediate result,
convolving the intermediate result by an analysis window having a temporal
width corresponding to an impulse response time of a low-pass filter to
produce X cos and X sin,
computing the phase data as arctan(X sin/X cos) when X cos>0 or computing
the phase data as arctan(X sin/X cos)+.pi. when X cos<0, and
computing the amplitude data as a square root of a sum of squares of X cos
and X sin.
9. A method as recited in claim 6, the step of adjusting a temporal change
rate of the periodically changing amplitude and frequency data for each of
the plurality of frequency band-divided waveform data comprising the steps
of:
duplicating a portion of the amplitude data for an amount equal to at least
one integer period of the amplitude data, and adding or omitting the
duplicated amplitude data to the amplitude data; and
duplicating a portion of the frequency data for an amount equal to at least
one integer period of the periodically changing frequency data, and adding
or omitting the duplicated frequency data to the frequency data.
10. A method as recited in claim 6, the steps of reproducing the amplitude
and frequency data and combining the reproduced amplitude and frequency
data for each of the plurality of frequency band-divided waveform data
comprising the steps of:
for each of the plurality of frequency band-divided waveform data,
resetting phases of the phase and frequency data extracted from the
frequency band-divided waveform data,
generating cosine waves in which a center frequency of the frequency
band-divided waveform data has been changed temporally in accordance with
the periodically changing amplitude of the frequency data, and
amplitude modulating the cosine waves with the amplitude data; and
combining the amplitude modulated cosine waves generated from the plurality
of frequency band-divided waveform data.
11. A method as recited in claim 6, the step of adjusting a temporal change
rate of the periodically changing amplitude and frequency data for each of
the plurality of frequency band-divided waveform data, comprising the
steps of:
for each of the plurality of frequency band-divided waveform data,
appending segment marks to the amplitude data that define segments of
amplitude data, and
duplicating a portion of the amplitude data beginning at a segment mark for
an amount equal to at least one integer period of the periodically
changing amplitude data, and adding or omitting the duplicated amplitude
data to the amplitude data.
12. A method as recited in claim 11, further including the step of
cross-fade processing to obtain a smooth temporal change rate of the
amplitude data.
13. A method as recited in claim 11, further including the step of
interpolation mutually adjoining portions of segments of the amplitude
data after adding or omitting the duplicated amplitude data to obtain a
smooth temporal change rate of the amplitude data.
14. A method for generating a compressed or expanded waveform from original
waveform data, the method comprising the steps of:
generating a plurality of frequency band-divided waveform data from the
original waveform data;
extracting frequency and amplitude data having periodically changing
amplitudes for each of the plurality of frequency band-divided waveform
data;
adjusting a temporal change rate of the periodically changing amplitude and
frequency data for each of the plurality of frequency band-divided
waveform data;
reproducing the amplitude and frequency data in accordance with the
adjusted temporal change rates of the amplitude and frequency data for
each of the plurality of frequency band-divided waveform data; and
combining the reproduced amplitude and frequency data for each of the
plurality of frequency band-divided waveform data to generate the
compressed or expanded waveform.
15. A method as recited in claim 14, the step of generating a plurality of
frequency band-divided waveform data comprising dividing the original
waveform data into a plurality of frequency band-divided waveform data so
that each frequency band contains a center frequency and a plurality of
harmonics that are integer multiples of the center frequency.
16. A method as recited in claim 15, the steps of dividing the original
waveform data into a plurality of frequency band-divided waveform data and
extracting frequency and amplitude data having periodically changing
amplitudes comprising the steps of:
for each of the plurality of frequency band-divided waveform data, and for
each of a plurality of N instantaneous values of the original waveform
data and a corresponding Nth data item of the center frequency received in
time sequence within each of the plurality of frequency band-divided
waveform data,
multiplying the Nth data item of the center frequency within a particular
frequency band-divided waveform data by the Nth instantaneous value of the
original waveform data to produce an intermediate result,
convolving the intermediate result by an analysis window having a temporal
width corresponding to an impulse response time of a low-pass filter to
produce X cos and X sin,
computing the frequency data by phase differentiating X cos and X sin, and
computing the amplitude data as a square root of a sum of squares of X cos
and X sin.
17. A method as recited in claim 14, the step of adjusting a temporal
change rate of the periodically changing amplitude and frequency data for
each of the plurality of frequency band-divided waveform data comprising
the steps of:
duplicating a portion of the amplitude data for an amount equal to at least
one integer period of the amplitude data, and adding or omitting the
duplicated amplitude data to the amplitude data; and
duplicating a portion of the frequency data for an amount equal to at least
one integer period of the periodically changing frequency data, and adding
or omitting the duplicated frequency data to the frequency data.
18. A method as recited in claim 14, the steps of reproducing the amplitude
and frequency data and combining the reproduced amplitude and frequency
data for the plurality of frequency band-divided waveform data comprising
the steps of:
for each of the plurality of frequency band-divided waveform data,
generating cosine waves in which a center frequency of the frequency
band-divided waveform data has been changed temporally in accordance with
the periodically changing amplitude of the frequency data, and
amplitude modulating the cosine waves with the amplitude data; and
combining the amplitude modulated cosine waves generated from the plurality
of frequency band-divided waveform data.
19. A method as recited in claim 14, further including the step of:
for each of the plurality of frequency band-divided waveform data, changing
pitch by
adding the frequency data to center frequency data to produce intermediate
frequency data,
multiplying the intermediate frequency data by a frequency change ratio to
produce new frequency data, and
duplicating a portion of the new frequency data for an amount equal to at
least one integer period of the new frequency data, and adding or omitting
the duplicated new frequency data to the new frequency data.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
Embodiments of the present invention relate to and claim priority to
Japanese Patent Application No. 11-254569, filed on Sep. 8, 1999, and the
contents of that application are incorporated by reference herein.
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a waveform reproduction apparatus with
which a waveform that has been compressed or expanded in the direction of
the temporal axis is reproduced.
2. Description of the Related Art
For some time, waveform reproduction apparatuses with which a waveform that
has been compressed or expanded in the direction of the temporal axis is
reproduced have been known. For these waveform reproduction apparatuses, a
number of formats have been proposed. Here an explanation will be given
first regarding a waveform reproduction apparatus that uses a cross-fade
format.
FIG. 12 is an explanatory diagram of a cross-fade format in which a musical
tone is compressed or expanded in the direction of the temporal axis.
In the waveform reproduction apparatus that uses the cross-fade format, the
waveform data that express the waveform of the musical tone are stored in
a RAM that is not shown in the diagram. The waveform data that have been
stored in the RAM are read out and, as is shown in FIG. 12(a), the
waveform data in a specified segment (known as the "opened segment") are
jump read and the waveform is compressed or, as is shown in FIG. 12(b),
the waveform data in a specified segment (known as the "repeated segment")
are repeated and read out and the waveform is expanded. By means of
carrying out this action, it is possible to restrain the changes in the
pitch with a waveform that has been compressed or expanded and to preserve
the pitch of the musical tone. In addition, with the cross-fade format,
since noise that is generated in the vicinity of the discontinuous areas
of the links between a particular segment and the segments that adjoin
that segment is suppressed, it is possible to carry out cross-fade
processing in the vicinity of the discontinuous areas.
Here, the meaning of cross-fade processing is processing in which, by means
of gradually increasing the amplitude of a waveform that has begun to be
read out anew (this is made the later waveform) together with the gradual
reduction of the amplitude of the waveform that has been read out up to
that point (this is made the head waveform), a transition is made smoothly
from the head waveform to the later waveform.
However, with this cross-fade format, since the waveforms that represent
musical tone waveforms that are continuous are jump read out or
repetitively read out directly, even though cross-fade processing is
carried out, there is a problem in that fluctuations or ripples are
produced in the waveform that has been compressed or expanded due to such
things as a shift in the phase.
In order to solve this problem, a waveform reproduction apparatus called a
phase vocoder has been presented. Below, an explanation will be given in
regular sequence regarding this phase vocoder.
With the phase vocoder, the original waveform, which expresses the original
musical tone prior to carrying out compression or expansion is input. The
phase vocoder divides the original waveform that has been input into a
multiple number of frequency bands.
FIG. 13 is a diagram that shows the multiple number of frequency bands that
have been divided by a phase vocoder.
The original waveform that has been input is divided into a multiple number
(here, there are 100) of frequency bands (band 0, 1, . . . k, . . . , p, .
. . , 99) which have the center frequencies .omega.0, .omega.1, . . . ,
.omega.k, . . . , .omega.p, . . . , .omega.99 that are respectively the
integer multiple frequencies that represent the fundamental frequency and
the harmonics of the fundamental frequency including the second harmonic,
third harmonic etc. In addition, this phase vocoder, for each waveform
component of the respective multiple number of frequency bands that have
been divided, extracts the frequency data and the amplitude data of each
of the waveform components that represent the frequencies that change in
order together with the passage of time (known as the instantaneous
frequency) and the amplitudes that change in order together with the
passage of time. The frequency data and the amplitude data that have been
extracted in this manner are stored in the memory.
At the time of the reproduction of the waveform, the temporal change rates
are adjusted for the frequencies and amplitudes that are expressed by the
frequency data and the amplitude data that have been extracted in each
frequency band.
FIG. 14 is a schematic diagram that shows the aspects of the frequency and
amplitude temporal change rates that have been adjusted by the phase
vocoder.
In FIG. 14(a), the amplitude envelope and the frequency envelope that are
expressed by the amplitude data and the frequency data that change
together with the passage of time in a certain single frequency band are
shown. As is shown in FIG. 14(b), the amplitude data and the frequency
data are corrected by the adjustment of the temporal change rate for the
frequency and the amplitude in accordance with the degree to which
expansion or compression are carried out and the envelope is expanded or,
as is shown in FIG. 14(c), the amplitude data and the frequency data are
culled out and the compression of the envelopes is carried out. By doing
it in this manner, after the amplitude envelopes and the frequency
envelopes have been adjusted for each frequency band, the cosine waves
that have been finely adjusted by an oscillator with which the fine
adjustment of the frequency is possible in accordance with the frequency
envelope for the center frequency of each of the frequency bands together
with the passage of time are obtained. The amplitudes of the cosine waves
are finely adjusted in accordance with the amplitude envelopes together
with the passage of time and, in addition, in this phase vocoder, all of
these waveforms that have been reproduced are combined. In this manner, a
reproduced waveform in which the original waveform that has been input has
been compressed or expanded in the direction of the temporal axis is
obtained.
Since the phase vocoder that has been discussed above is one in which the
original waveform is divided into a multiple number of frequency bands,
the temporal change rates of the frequencies and the amplitudes that
change together with the passage of time are adjusted for each of multiple
number of frequency bands that have been divided and, by means of the
reproduction of the time conversions for the frequencies and the
amplitudes following adjustment, a reproduced waveform in which the
original waveform has been compressed or expanded in the direction of the
temporal axis is obtained, compared to the case, as in the waveform
reproduction apparatus that uses the cross-fade format, in which the
waveform data that express the original waveform are themselves directly
jump read out or repetitively read out, noise and fluctuations due to such
things as a shift in the phase are reduced.
However, in this phase vocoder, with such things as voices and brass where
the period of the waveform is long or the waveforms of chords, if the
expansion and compression rate, which represents the proportion of
compression or expansion, is varied greatly from 1.0 for both compression
and expansion, there is a breakdown of the harmonic relationships of the
musical tones that are expressed by the waveforms that have been
compressed or expanded in the direction of the temporal axis. A detailed
explanation of this phenomenon will be given below.
In the case of the phase vocoder discussed above, in order to provide a
theoretical description, it was explained to the effect that the original
waveform that has been input is, as is shown in FIG. 12, divided into a
frequency band that contains the fundamental frequency, a frequency band
that contains only a frequency that is twice the fundamental frequency
etc. and frequency bands that contain one each only from among the
multiple number of frequency components that comprise the original
waveform in a single frequency band. However, in the case of this kind of
method of division, the requirement is produced for a division into an
extremely large number of frequency bands, an extremely large circuit
becomes necessary or the time needed for the operations becomes extremely
long and it is not pragmatic. Therefore, here, the division of the
frequency bands such that a multiple number of frequency components that
comprise the original waveform are contained in a single frequency band is
considered.
FIG. 15 is a diagram that shows a multiple number of frequency bands and
FIG. 16 is a diagram that shows the shape of the pulse stream form
original waveform prior to the division into the multiple number of
frequency bands that are shown in FIG. 15. In addition, FIG. 17 is a
diagram that shows the waveform in a single frequency band from among the
multiple number of frequency bands that are shown in FIG. 15.
Here, as is shown in FIG. 16, the original waveform that is input into the
phase vocoder comprises a periodic pulse stream that has a comparatively
long period. The number of band divisions that are shown in FIG. 15 is
smaller than the number of band division that are shown in FIG. 13 and,
consequently, the bandwidths for each individual frequency band are wide.
Because of this, as is shown in FIG. 15, in, for example, band k, which is
one divided band, a multiple number of frequencies which are integer
multiples of the fundamental frequency that corresponds to the fundamental
period exist that represent a multiple number of adjoining harmonics. The
waveform in this band k is the waveform that is shown by the solid line in
FIG. 17 and, as is shown by the broken like that represents the envelope,
is a waveform that is amplitude modulated at the fundamental period T.
FIG. 18 and FIG. 19 are diagrams that show the aspects of the waveform
components in band k that is shown in FIG. 17 in which the temporal change
rates are adjusted so that the amplitude and the frequency change slowly.
In addition, FIG. 20 is a diagram that shows the waveforms in band k after
the temporal change rates of the amplitude and the frequency have been
adjusted so that they are slow.
The broken lines a and b that are shown in FIG. 18 and FIG. 19 are the
envelopes prior to the adjustment of the temporal change rates of the
amplitude and the frequency in band k. In adjusting the temporal change
rates of the amplitude and the frequency in band k so they are slow, the
amplitude data and the frequency data of each envelope that is shown by
the broken lines a and b at each sampling point are interpolated uniformly
in the direction of the temporal axis and are expanded as is shown by the
solid lines A and B. In this manner, the waveform that is shown in FIG. 20
in which the temporal change rates of the amplitude and the frequency of
band k are adjusted so that they are slow is obtained. Here, the
fundamental period T' of the waveform that is shown in FIG. 20 is longer
than the fundamental period of the waveform T that is shown in FIG. 17.
When these kinds of waveforms are reproduced for each band and combined
and a waveform that has been expanded in the direction of the temporal
axis is obtained, there is a problem in that the harmonic relationships of
the original waveform are lost and the sound quality of the musical tone
is lowered. In order to avoid that, it is necessary that the original
waveform that is input, as is shown in FIG. 13, be divided into many
frequency bands in which the fundamental frequency and the frequencies
that are integer multiples of the fundamental frequency are the center
frequencies. However, when they are divided into a large number of
frequency bands in this manner, as was discussed before, the amount of
processing in the phase vocoder swells, the processing time becomes longer
together with an increase in the size of the circuit and, consequently, as
a practical matter, the realization of the system becomes difficult.
In addition, with the phase vocoder of the past that has been described
above, the reproduction of the original sound (hereafter, referred to as
"one-to-one reproduction") is carried out again and again. In that case,
the temporal change rate of the frequency and the amplitude and the pitch
data are adjusted so that neither compression nor expansion in the
direction of the temporal axis is carried out for each of the multiple
number of frequency bands of the original waveform that have been divided
and one-to-one reproduction can be carried out. However, the phase data
are not taken into consideration. Because of this, with one-to-one
reproduction, waveforms are reproduced that have phases that are different
from the phase of the waveform that expresses the original sound and,
consequently, there are problems such as the fact that the tone quality is
degraded and the orientation of the stereo signal is lost.
SUMMARY OF THE INVENTION
The present invention, taking the above mentioned conditions into
consideration, has as its object the presentation of a waveform
reproduction apparatus with which a waveform that has been compressed or
expanded in the direction of the temporal axis can be obtained that
expresses such things as musical tones the sound quality of which is high.
The first waveform reproduction apparatus from among the waveform
reproduction apparatuses of the present invention that achieves the above
mentioned object is characterized in that it comprises a storage means in
which, for each waveform component at the time that an original waveform
is divided into each waveform component of a multiple number of frequency
bands, the phase data and the amplitude data that respectively express the
phases and amplitudes of each waveform component that change in order
together with the passage time are stored and a frequency data conversion
means in which the above mentioned phase data are converted into frequency
data and a change rate adjustment means in which the temporal change rates
of the frequency and the amplitude that are expressed by the above
mentioned phase data and amplitude data are adjusted and a waveform
reproduction means in which a waveform in which the original waveform has
been compressed or expanded in the direction of the temporal axis is
obtained by the reproduction of a waveform in which the time conversions
of the frequency and the amplitude following the adjustments of the
temporal change rate are reproduced.
With the first waveform reproduction apparatus of the present invention,
since the phase data and the amplitude data for each waveform component of
the original waveform are stored in advance and a waveform is obtained in
which the original waveform has been compressed or expanded in the
direction of the temporal axis by adjusting the temporal change rate of
the frequency and the amplitude that are expressed by the phase data and
amplitude data that have been stored and reproducing the waveform, even in
a case where one-to-one reproduction is carried out in order to reproduce
the original sound, for each of the multiple number of frequency bands of
the original waveform that have been divided, the temporal change rates of
the frequency and the amplitude that are expressed by the phase data and
the amplitude data are adjusted so that neither compression nor expansion
are carried out in the direction of the temporal axis. Accordingly,
compared to the technology with which a waveform is reproduced that has a
phase that is different from the phase of the waveform that expresses the
original sound, there are no problems such as a degradation of the tone
quality or a loss of the orientation of the stereo signal and it is
possible to obtain a waveform that has been compressed or expanded in the
direction of the temporal axis which expresses such things as musical
tones the sound quality of which is high.
Here, it is effective if the above mentioned change rate adjustment means
is one, with regard to the amplitude, in which, by means of an operation
in which the amplitude data for the amount of one integer period or more
of the periodic change of the amplitude that is expressed by the amplitude
data are duplicated and added or omitted, the temporal change rate of the
amplitude is adjusted so that it is longer than that period while
maintaining the period of the periodic change of the amplitude.
When, in this manner, while maintaining the period of the periodic change
of the amplitude, the time conversion of the amplitude is adjusted so that
it is longer than that period, even in those cases where, in a certain
frequency band, a multiple number of harmonics that are adjacent exist in
multiple numbers, without compressing or expanding the fundamental
frequency of the waveform components that are in the frequency band, it is
possible to prevent the breakdown of the harmonic relationships of the
original waveform and to raise the sound quality of the musical tones etc.
In addition, it is preferable that the above mentioned change rate
adjustment means be one in which, together with the adjustment of the
temporal change rate for the amplitude by the repetition of the above
mentioned operations related to the amplitude data, with regard to the
frequency that is expressed by the phase data also, the temporal change
rate for the frequency is adjusted by repeating an operation in which the
phase data that correspond to the amplitude data that are duplicated and
added or omitted or the frequency data into which that phase data have
been transformed are repeated and added or omitted.
When, in this manner, with regard to the frequency that is expressed by the
phase data also, the operation in which the phase data that correspond to
the amplitude data that are duplicated and added or omitted or the
frequency data into which that phase data have been transformed are
duplicated and added or omitted, it is possible to adjust the temporal
change rate of the frequency with good accuracy.
In addition, the second waveform reproduction apparatus from among the
waveform reproduction apparatuses of the present invention that achieves
the above mentioned object is characterized in that it comprises a storage
means in which, for each waveform component at the time that an original
waveform is divided into each waveform component of a multiple number of
frequency bands, the frequency data and the amplitude data that
respectively express the frequencies and amplitudes of each waveform
component that change in order together with the passage time are stored.
The second waveform reproduction apparatus also comprises a change rate
adjustment means in which the temporal change rates of the frequency and
the amplitude that are expressed by the above mentioned frequency data and
amplitude data are adjusted. In addition, the second waveform reproduction
apparatus comprises a waveform reproduction means in which a waveform in
which the original waveform has been compressed or expanded in the
direction of the temporal axis is obtained by the reproduction of a
waveform in which the time conversions of the frequency and the amplitude
following the adjustments of the temporal change rate are reproduced.
The above mentioned change rate adjustment means is one, with regard to the
amplitude, in which, by means of an operation in which the amplitude data
for the amount of one integer period or more of the periodic change of the
amplitude that is expressed by the amplitude data are duplicated and added
or omitted, the temporal change rate of the amplitude is adjusted so that
it is longer than that period while maintaining the period of the periodic
change of the amplitude.
Since, with the second waveform reproduction apparatus of the present
invention, the period of the periodic change of the amplitude is
maintained and the temporal change rate of the amplitude is adjusted so
that the period is longer than that period, even in those cases where, in
a certain frequency band, a multiple number of harmonics that are adjacent
exist in multiple numbers, without compressing or expanding the
fundamental frequency of the waveform components that are in the frequency
band, it is possible to prevent the breakdown of the harmonic
relationships of the original waveform and to raise the sound quality of
the musical tones etc.
Here, it is preferable that the above mentioned change rate adjustment
means be one in which, together with the adjustment of the temporal change
rate for the amplitude by the repetition of the above mentioned operations
related to the amplitude data, with regard to the frequency also, the
temporal change rate for the frequency is adjusted by repeating an
operation in which the frequency data that correspond to the amplitude
data that are duplicated and added or omitted are repeated and added or
omitted.
When, in this manner, with regard to the frequency also, the operation in
which the frequency data that correspond to the amplitude data that are
duplicated and added or omitted are duplicated and added or omitted, it is
possible to adjust the temporal change rate of the frequency with good
accuracy.
As has been explained above, in accordance with the present invention, it
is possible to obtain a waveform that expresses musical tones etc. that
have a high sound quality which have been compressed or expanded in the
direction of the temporal axis.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a block diagram that shows the circuit configuration of the
waveform reproduction apparatus of the first preferred embodiment of the
present invention.
FIG. 2 is a block diagram in which the functions of the RAM and the DSP
that are shown in FIG. 2 have been shown as functional blocks and in which
the blocks are shown classified by function.
FIG. 3 is a diagram that shows the waveform processing in the first channel
that comprises the analysis section that is shown in FIG. 2.
FIG. 4 is a diagram that shows the aspect of the adjustment of the temporal
change rate by the time and frequency conversion processing means 220_k
that is comprised by the change section 220 so that the amplitude in band
k changes slowly.
FIG. 5 is a diagram that shows the aspect of the adjustment of the temporal
change rate by the time and frequency conversion processing means 220_k
that is comprised by the change section 220 so that the frequency in band
k changes slowly.
FIG. 6 is a diagram that shows the time and frequency conversion processing
in order to change the sound pitch in the time and frequency conversion
processing circuit.
FIG. 7 is a diagram that shows the segment mark that has been established
for the adjustment of the temporal change rate for the amplitude in band
k.
FIG. 8 is a diagram that shows the aspect in which a smooth amplitude
envelope is obtained by the interpolation of the mutually adjoining
portions of segments after the omission or addition of the opened segment
has been carried out.
FIG. 9 is a block diagram in which the functions of the RAM and the DSP
that are comprised by the waveform reproduction apparatus of the second
preferred embodiment of the present invention have been shown as
functional blocks and in which the blocks are shown classified by
function.
FIG. 10 is a diagram that shows the waveform processing in the first
channel that comprises the analysis section that is shown in FIG. 9.
FIG. 11 is a diagram that shows the frequency conversion processing in
order to change the sound pitch in the time and frequency conversion
processing means that is shown in FIG. 9.
FIG. 12 is an explanatory diagram of the cross-fade format in which the
waveform of the musical tone is compressed or expanded in the direction of
the temporal axis.
FIG. 13 is a diagram that shows the multiple number of frequency bands that
have been divided by the phase vocoder.
FIG. 14 is a schematic diagram that shows the aspect in which the temporal
change rates of the frequency and the amplitude are adjusted by the phase
vocoder.
FIG. 15 is a diagram that shows a multiple number of frequency bands.
FIG. 16 is a diagram that shows a pulse stream form original waveform prior
to division into the multiple number of frequency bands that are shown in
FIG. 12.
FIG. 17 is a diagram that shows one frequency band from among the multiple
number of frequency bands that are shown in FIG. 15.
FIG. 18 is a diagram that shows the aspect of the adjustment of the
temporal change rate so that the amplitude of the waveform component in
band k that is shown in FIG. 17 changes slowly.
FIG. 19 is a diagram that shows the aspect of the adjustment of the
temporal change rate so that the frequency of the waveform component in
band k that is shown in FIG. 17 changes slowly.
FIG. 20 is a diagram that shows the waveform after the temporal change
rates of the amplitude and frequency in band k have been adjusted so that
they are slow.
DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
Below, an explanation will be given regarding preferred embodiments of the
present invention.
FIG. 1 is a block diagram that shows the circuit configuration of the
waveform reproduction apparatus of the first preferred embodiment of the
present invention.
The waveform reproduction apparatus 100 comprises the CPU 10 and the
digital signal processor (DSP) 20. The DSP 20 is controlled by the CPU 10
and, as will be discussed later, forms a waveform that has been compressed
or expanded.
In addition, the waveform reproduction apparatus 100 comprises the ROM 30,
the first RAM 40 and the operator group 50. The programs to carry out the
operation of the CPU 10 and the DSP 20 are stored in the ROM 30 and the
program for the DSP 20 is transmitted to the DSP 20 via the CPU 10. The
first RAM 40 is used as the working memory for the CPU 10. The operator
group 50 comprises the expansion and compression rate switch for setting
the expansion and compression rate and the reproduction switch for the
combination and reproduction of each of the waveforms that have been
formed based on the expansion and compression rates that have been set.
In addition, the waveform reproduction apparatus 100 comprises the A/D
converter 60, the second RAM 70 and the D/A converter 80. The A/D
converter 60 converts the analog signal A that has been input into a
digital one, forms the digital original waveform x(n) and inputs it to the
DSP 20. Such things as the phase data and the amplitude data that have
been produced by the DSP 20, as will be discussed later, are stored in the
second RAM 70. The D/A converter 80 converts the compressed or expanded
waveform y(n) that has been output by the DSP 20 into an analog one and
outputs the analog signal B.
FIG. 2 is a block diagram in which the functions of the RAM and the DSP
that are shown in FIG. 2 have been shown as functional blocks and in which
the blocks are shown classified by function.
The DSP 20 that is shown in FIG. 2 comprises the analysis section 210, the
conversion section 220 and the combining section 230. The expansion and
compression switch (not shown in the figure) that is comprised by the
operator group 50 of the waveform reproduction apparatus 100 that is shown
in FIG. 1 is operated and the desired rate of expansion or compression is
set. Then the reproduction switch, which is not shown in the figure, is
pressed directing the waveform reproduction and the original waveform x(n)
that expresses the original musical tone prior to carrying out compression
or expansion is input to the analysis section 210 that is comprised by the
waveform reproduction apparatus 100. Here, n is a number that has been
appended to each piece of data that expresses the instantaneous value of
the original waveform which are input in order in time sequence. The
analysis section 210 comprises the channels 210_0, 210_1, . . . , 210_k, .
. . , 210_p. In the analysis section 210, the original waveform x(n) that
has been input, as is shown in FIG. 15 that was discussed previously, is
divided into a multiple number of frequency bands (band 0, 1, . . . , k, .
. . , p) so that, in each band, the frequencies that express the multiple
number of harmonics that are adjacent and are integer multiples of the
fundamental frequency exist in multiple numbers. For each of the waveform
components of the multiple number of frequency bands that have been
divided, the phase data and amplitude data that express the respective
phases and amplitudes that have changed in order together with the passage
of time are extracted and output to the RAM 70. A detailed explanation
will be given below regarding the particulars of the analysis section 210
referring to FIG. 3.
FIG. 3 is a diagram that shows the waveform processing in the first channel
that comprises the analysis section that is shown in FIG. 2.
In FIG. 3, the aspect of the waveform processing in channel 210_k, which
represents the multiple number of channels, is shown. In this channel
210_k, the nth data item (cos(.omega.kn), sin(.omega.kn)) of the center
frequency .omega.k for the frequency band (band k) that corresponds to
that channel 210_k is multiplied with the original waveform x(n) that has
been input and converted into a real number portion and an imaginary
number portion. Next, it is opened or convolved by the analysis window w
that has a temporal width that corresponds to the impulse response time of
an equivalent analog low-pass filter. Then the calculation of arctan(x
sin/x cos), when x cos >0, is carried out, the calculation of arctan(x
sin/x cos)+.pi., when x cos<0, is carried out and, together with the
extraction of the phase data, the amplitude data are extracted by deriving
the square root of the sum of the squares. This kind of calculation is
repeated while the n of the nth data item (cos(.omega.kn, sin(.omega.kn))
of the center frequency .omega.k and the analysis window w(n) advance in
order, that is to say, as they advance in order temporally. In this
manner, the phase data and amplitude data that express the respective
phases and amplitudes that have changed in order together with the passage
of time are extracted for each of the channels 210_0, 210_1, . . . ,
210_k, . . . , 201_p that are comprised by the analysis section 210. The
phase data and the amplitude data that have been extracted are stored in
the RAM 70 that is shown in FIG. 2.
The phase data and the amplitude data that have been stored in the RAM 70
are input to the conversion section 220. The conversion section, as is
shown in FIG. 2, comprises the multiple number of time and frequency
conversion processing means 220_0, 220_1, . . . , 220_k, . . . , 220_p.
Each of the time and frequency conversion processing means 220_0, 220_1, .
. . , 220_k . . . , 220_p converts the phase data into frequency data for
each frequency band. In addition, each time and frequency conversion
processing means 220_0, 220_1, . . . , 220 k, . . . , 220_p, for each
frequency band, with regard to the amplitude, by means of a periodic
operation in which the amplitude data for the amount of one integer period
or more of the periodic change of the amplitude that is expressed by the
amplitude data is duplicated and added or omitted, the temporal change
rate of the amplitude is adjusted while maintaining the period of the
periodic change of the amplitude.
Thus, for each frequency band, the amplitude data is examined to see if
there is any cyclic amplitude modulation present. Such an amplitude
modulation is shown in FIG. 4. If such a modulation is present, then the
duration of the modulation cycle, indicated by the duration between
vertical lines in FIG. 4, is used as the period of duplication and
addition or omission of phase and amplitude envelope data to effect time
expansion or compression. If no such modulation is present, then a
convenient duplication period is used to effect the time compression or
expansion.
In addition, with regard to the frequency that is expressed by the phase
data also, the temporal change rate of the frequency is adjusted by means
of repeating an operation in which the phase data that correspond to the
amplitude data that are duplicated and added or omitted or the frequency
data into which that phase data have been transformed are duplicated and
added or omitted.
FIG. 4 and FIG. 5 are diagrams that show the aspect of the adjustment of
the temporal change rate by the time and frequency conversion processing
means 220_k that is comprised by the change section 220 so that the
amplitude and frequency respectively in band k change slowly.
The thin lines a and b that are shown in FIG. 4 and FIG. 5 are,
respectively the envelopes of the amplitude and the frequency that is
expressed by the phase data in band k prior to the adjustment of the
temporal change rates. Incidentally, FIG. 5 originally showed phase data
but, for the purpose of making it easier to understand, it now shows the
frequency data that are expressed by the phase data. To adjust the
temporal change rate of the amplitude in band k so that it becomes slow,
the amplitude data for the amount of one period of the periodic change of
the amplitude that the amplitude data indicated by the thin line a
expresses are duplicated and added as shown by the thick line a. In
addition, with regard to the frequency that is expressed by the phase
data, the frequency data into which the phase data that correspond to the
amplitude data that have been duplicated and added are duplicated and
added as shown by the thick line B. In this manner, together with the
expansion of the amplitude envelope in band k, the frequency envelope is
also expanded while maintaining the period of the periodic change of the
amplitude. Incidentally, it is not limited to this and it may be done so
that the amplitude data for the amount of two or more integer periods are
duplicated and added or omitted and, with regard to the frequency also,
together with the further expansion or compression of the amplitude
envelope in band k, the frequency data that correspond to the amplitude
data that are duplicated and added or omitted are duplicated and added or
omitted and the frequency envelope in band k is further expanded or
compressed.
Incidentally, in those cases where the pitch of the sound is changed, the
frequency conversion processing that is shown below may be carried out by
the time and frequency conversion processing means prior to the adjustment
of the temporal change rate of the frequency.
FIG. 6 is a diagram that shows the time and frequency conversion processing
in order to change the sound pitch in the time and frequency conversion
processing circuit.
The time and frequency conversion processing means comprises a read-out
means in which the amplitude data and the phase data are input from the
RAM 70. As is shown in FIG. 4 and FIG. 5 that were discussed previously
and in FIG. 7 that will be discussed later, an extension processing in
which the adjustment of the time change rate is repeated is carried out by
this read-out means. In order to carry out the frequency conversion
processing to change the pitch of the sound, the phase data that have been
output from the read-out means are differentiated by the time and
frequency conversion processing means and the frequency data are
extracted. Since these frequency data are data that only vary in the
frequency band that corresponds to that time and frequency conversion
processing means, the frequency are added to the center frequency data of
the band by the time and frequency conversion processing means and the
frequency data that include the data for the center frequency of the band
are obtained. Then these are multiplied by the frequency conversion ratio
that has been established in advance and new frequency data are obtained.
By means of the adjustment of the temporal change rate of the frequency
based on these new frequency data, as was explained referring to FIG. 5,
it is possible to obtain a frequency envelope in which the pitch of the
sound has been changed.
FIG. 7 is a diagram that shows the segment mark that has been established
for the adjustment of the temporal change rate for the amplitude in band
k. Segment marks may be established at zero-crossings of the amplitude
data, and define portions of the waveform to be duplicated and added or
omitted.
In order to append the segment marks for the adjustment of the temporal
change rate of the amplitude to the amplitude envelope in band k, each of
the points at which each segment mark is to be appended is calculated in
advance based on the amplitude data that have been stored in the RAM 70
and the data that indicate each point are stored in the RAM 70 together
with the amplitude data. Having done this, afterward at the time that the
temporal change rate of the amplitude in band k is adjusted, those data
are read out and, as is shown in FIG. 7(a), the waveform in the segment
that has been opened is omitted and the amplitude envelope of band k is
compressed or, as is shown in FIG. 7(b), the waveform in the segment that
has been opened is duplicated and added and the amplitude envelope of band
k is expanded. In this preferred embodiment, the amplitude data for the
amount of one integer period or more of the periodic change of the
amplitude that is expressed by the amplitude data are duplicated or
omitted but there are cases where it is not a completely repeated waveform
and the waveforms in the duplicated or omitted portions do not connect
well. Therefore, in FIG. 7(a) and FIG. 7(b), a smooth amplitude envelope
is obtained by the cross-fade processing of the mutually adjacent portions
after the omission or addition of the opened segments has been carried
out.
Next, an explanation will be given regarding a method with which a smooth
amplitude envelope is obtained that can be employed instead of the
cross-fade processing.
FIG. 8 is a diagram that shows the aspect in which a smooth amplitude
envelope is obtained by the interpolation of the mutually adjoining
portions of segments after the omission or addition of the opened segment
has been carried out.
In FIG. 8, the segment a and the segment b after the omission or addition
of the opened segment has been carried out are shown. The mutually
adjacent portions of the segment a and the segment b may be interpolated
by an interpolation means (not shown in the diagram) and connected as with
the broken line c obtaining a smooth amplitude envelope.
The amplitude data, the frequency data and the phase data that express the
temporal changes of the amplitudes and frequencies after the temporal
change rates have been adjusted for each frequency band in the above
manner are input to the combining section 230 from the conversion section
220 that is shown in FIG. 2. In addition, the phase reset signals from a
circuit section that is not shown in the drawing of the DSP 20 are input
to the combining section 230. The combining section 230, as is shown in
FIG. 2, comprises the cosine signal generator 230_0 and modulator 231_0
pair, the cosine signal generator 230_1 and modulator 231_1 pair, . . . ,
the cosine signal generator 230_k and modulator 231_k pair, . . . and the
cosine signal generator 230_p and modulator 231_p pair. The phase reset
signals and the frequency data and phase data from the time and frequency
conversion means 220_0, 220_1, . . . , 220.sub.13 k, . . . , 220_p are
respectively input to the cosine signal generators 230_0, 230_1, . . . ,
230_k, . . . , 230_p. These cosine signal generators 230_0, 230_1, . . . ,
230_k, . . . , 230_p reset the phases that are being maintained when the
phase reset signals are input, acquire the phase data from the time and
frequency conversion means 220_0, 220_1, . . . , 220_k, . . . , 220_p and
rewrite them with the value of the center frequency that has had the
portion of the rotation .omega.kn added. Incidentally, the phase reset
signals are input only once at the start of reproduction. In addition, the
cosine signal generators 230_0, 230_1, . . . , 230_k, . . . , 230_p
generate cosine waves in which the center frequency of each of the
frequency bands has been changed temporally in accordance with the
frequency envelopes that are expressed by each of the groups of frequency
data. These cosine waves are input to each of the respective modulators
231_0, 231_1, . . . , 231_k, . . . , 231_p. In addition, each of the
groups of amplitude data from each of the time and frequency conversion
processing means 220_0, 220_1220_k, . . . , 220_p is input to each of the
modulators 231_0, 231_1, . . . , 231_k, . . . , 231_p. Each of the
modulators 231_0, 231_1, . . . , 231_k, . . . , 231_p amplitude modulates
each of the cosine waves from each of the cosine signal generators 230_0,
230_1, . . . , 230_k, . . . , 230_p with the amplitudes that are expressed
by each of the groups of amplitude data from each of the time and
frequency conversion means 220_0, 220_1, . . . , 220_k, . . . , 220_p. In
this manner, the waveforms in which the temporal changes of the frequency
and amplitude of each band have been reproduced after the temporal change
rates have been adjusted are reproduced. In addition, all of the these
waveforms that have been reproduced are combined in the combining section
230. In this manner, the waveform y(n) in which the original waveform that
has been input is compressed or expanded in the direction of the temporal
axis. Incidentally, since, with the cosine signal generators 230_0, 230_1,
. . . , 230_k, . . . 230_p, it is possible to acquire the phase data that
correspond to the reproduction time, even in those cases where waveform
reproduction is carried out in which waveforms that have been compressed
or expanded in direction of the temporal axis or the waveform reproduction
is carried out not from the beginning of original waveform but from the
middle of the original waveform, it is possible to carry out the
reproduction of a waveform that has a phase that is the same as the phase
of the middle portion of the original waveform.
Since with the waveform reproduction apparatus 100 of the first preferred
embodiment, the waveform y(n) is obtained by means of the processing as
above, compared to a waveform reproduction apparatus that employs a
cross-fade format with which the waveform data that express the original
waveform are directly jump read or repetitively read out and cross-fade
processed, such things as fluctuation and ripples due to a shift in the
phase that is produced in the vicinity of the discontinuous areas are
reduced.
In addition, with the waveform reproduction apparatus 100 of the first
preferred embodiment, even in those cases where something such as
one-to-one reproduction is carried out to reproduce the original sound,
the time change rates of the frequency and the amplitude that are
expressed by the phase data and the amplitude data are adjusted for each
of the multiple number of frequency bands into which the original waveform
has been divided so that neither compression nor expansion is carried out
in the direction of the temporal axis. Accordingly, a waveform is
reproduced that has the same phase as the phase of the waveform that
expresses the original sound and, compared to the technology of the past
with which a waveform is reproduced that has a phase that is different
from the phase of the waveform that expresses the original sound, there
are no problems such as a degradation of the tone quality or a loss of the
orientation of the stereo signal and it is possible to obtain a waveform
that has been compressed or expanded in the direction of the temporal axis
which expresses such things as musical tones the sound quality of which is
high.
Furthermore, since, with the waveform reproduction apparatus 100 of the
first preferred embodiment, the temporal change rate of the amplitude is
adjusted so that it is longer than that period while maintaining the
period of the periodic change of the amplitude, even in those cases where,
in a certain frequency band, the frequencies that express the multiple
number of harmonics that are adjacent and are integer multiples of the
fundamental frequency that corresponds to the fundamental period exist in
multiple numbers, without compressing or expanding the fundamental period,
it is possible to prevent the breakdown of the harmonic relationships of
the original waveform. Accordingly, it is possible to improve the sound
quality of musical tones etc. that are expressed by waveforms that have
been compressed or expanded in the direction of the temporal axis.
Next, an explanation will be given regarding a waveform reproduction
apparatus of the second preferred embodiment of the present invention.
Since the block diagram that shows the circuit configuration of the
waveform reproduction apparatus of the second preferred embodiment is the
same as the block diagram that shows the circuit configuration of the
waveform reproduction apparatus that is shown in FIG. 1 described
previously, it will be omitted from the explanation.
FIG. 9 is a block diagram in which the functions of the RAM and the DSP
that are comprised by the waveform reproduction apparatus of the second
preferred embodiment of the present invention have been shown as
functional blocks and in which the blocks are shown classified by
function.
The DSP 20 that is shown in FIG. 9 comprises the analysis section 210, the
conversion section 220 and the combining section 230. The expansion and
compression switch that is comprised by the operator group of the waveform
reproduction apparatus is operated and a desired expansion and compression
rate is set. Then, the reproduction switch is pressed, the waveform
reproduction is directed and the original waveform x(n) that expresses the
original musical tone prior to carrying out compression or expansion is
input to the analysis section 210. Here, n is a number that has been
appended to each piece of data that expresses the instantaneous value of
the original waveform which are input in order in time sequence. The
analysis section 210 comprises the channels 210_0, 210_1, . . . , 210_k, .
. . , 210_p. In the analysis section 210, the original waveform x(n) that
has been input, as is shown in FIG. 15 that was discussed previously, is
divided into a multiple number of frequency bands (band 0, 1, . . . , k, .
. . , p) so that, in each band, the frequencies that express the multiple
number of harmonics that are adjacent and are integer multiples of the
fundamental frequency exist in multiple numbers. For each of the waveform
components of the multiple number of frequency bands that have been
divided, the frequency data and amplitude data that express the respective
frequencies and amplitudes that have changed in order together with the
passage of time are extracted and output to the RAM 70. A detailed
explanation will be given below regarding the particulars of the analysis
section 210 referring to FIG. 10.
FIG. 10 is a diagram that shows the waveform processing in the first
channel that comprises the analysis section that is shown in FIG. 9.
In FIG. 10, the aspect of the waveform processing in channel 210_k, which
represents the multiple number of channels, is shown. In this channel
210_k, the nth data item (cos(.omega.kn, sin(.omega.kn)) of the center
frequency .omega.k for the frequency band (band k) that corresponds to
that channel 210_k is multiplied with the original waveform x(n) that has
been input and converted into a real number portion and an imaginary
number portion. Next, it is opened by the analysis window w that has a
temporal width that corresponds to the impulse response time of an
equivalent analog low-pass filter. Then, together with the phase
differentiation and the extraction of the frequency data, the amplitude
data are extracted by deriving the square root of the sum of the squares.
This kind of calculation is repeated while the n of the nth data item
(cos((.omega.kn, sin(.omega.kn)) of the center frequency .omega.k and the
analysis window w(n) advance in order, that is to say, as they advance in
order temporally. In this manner, the frequency data and amplitude data
that express the respective frequencies and amplitudes that have changed
in order together with the passage of time are extracted for each of the
channels 210_0, 210_1, . . . , 210_k, . . . , 210_p that are comprised by
the analysis section 210. The frequency data and the amplitude data that
have been extracted are stored in the RAM 70 that is shown in FIG. 9.
The phase data and the amplitude data that have been stored in the RAM 70
are input to the conversion section 220. The conversion section, as is
shown in FIG. 9, comprises the multiple number of time and frequency
conversion processing means 220_0, 220_1, . . . , 220_k, . . . , 220_p.
Each time and frequency conversion processing means 220_0, 220_1, . . . ,
220_k, . . . , 220_p, for each frequency band, with regard to the
amplitude, by means of a periodic operation in which the amplitude data
for the amount of one integer period or more of the periodic change of the
amplitude that is expressed by the amplitude data is duplicated and added
or omitted, the temporal change rate of the amplitude is adjusted so that
it is longer than that period while maintaining the period of the periodic
change of the amplitude. In addition, with regard to the frequency also,
the temporal change rate of the frequency is adjusted by means of
repeating an operation in which the frequency data that correspond to the
amplitude data that are duplicated and added or omitted are duplicated and
added or omitted.
In order to make the adjustment with the time and frequency conversion
processing means that is comprised by the conversion section 220 so that
the temporal change rate of the amplitude in band k becomes slow, the
amplitude data for the amount of one period of the periodic change of the
amplitude that the amplitude data indicated by the thin line a express are
duplicated and added as shown by the thick line a. In addition, with
regard to the frequency, as is shown in FIG. 5 which was discussed
previously, the frequency data that correspond to the amplitude data that
have been duplicated and added are duplicated and added as shown by the
thick line B. In this manner, together with the expansion of the amplitude
envelope in band k, the frequency envelope is also expanded while
maintaining the period of the periodic change of the amplitude.
Incidentally, it is not limited to this and it may be done so that the
amplitude data for the amount of two or more integer periods are
duplicated and added or omitted and, with regard to the frequency also,
together with the further expansion or compression of the amplitude
envelope in band k, the frequency data that correspond to the amplitude
data that are duplicated and added or omitted are duplicated and added or
omitted and the frequency envelope in band k is further expanded or
compressed.
Incidentally, in those cases where the pitch of the sound is changed, the
frequency conversion processing that is shown below is carried out by the
time and frequency conversion processing means prior to the adjustment of
the temporal change rate of the frequency.
FIG. 11 is a diagram that shows the frequency conversion processing in
order to change the sound pitch in the time and frequency conversion
processing means that is shown in FIG. 9.
In the time and frequency conversion processing means, the frequency data
are input from the RAM 70. Since these frequency data are data that only
vary in the frequency band that corresponds to that time and frequency
conversion processing means, the frequency are added to the center
frequency data of the band by the time and frequency conversion processing
means and the frequency data that include the data for the center
frequency of the band are obtained. Then these are multiplied by the
frequency change ratio that has been established in advance and new
frequency data are obtained. By means of the adjustment of the temporal
change rate of the frequency based on these new frequency data, as was
explained referring to FIG. 5, it is possible to obtain a frequency
envelope in which the pitch of the sound has been changed.
The frequency data and the amplitude data that express the temporal changes
of the amplitudes and frequencies after the temporal change rates have
been adjusted for each frequency band in the above manner are input to the
combining section 230 from the conversion section 220 that is shown in
FIG. 9. The combining section 230, as is shown in FIG. 9, comprises the
cosine signal generator 230_0 and modulator 231_0 pair, the cosine signal
generator 230_1 and modulator 231_1 pair, . . . , the cosine signal
generator 230_k and modulator 231_k pair, . . . and the cosine signal
generator 230_p and modulator 231_p pair. Each of the groups of frequency
data from the time and frequency conversion means 220_0, 220_1, . . . ,
220_k, . . . , 220_p are respectively input to the cosine signal
generators 230_0, 230_1, . . . , 230_k, . . . , 230_p. The cosine signal
generators 230_0, 230_1, . . . , 230_k, . . . , 230_p generate cosine
waves in which the center frequency of each of the frequency bands has
been changed temporally in accordance with the frequency envelopes that
are expressed by each of the groups of frequency data. These cosine waves
are input to each of the respective modulators 231_0, 231_1, . . . ,
231_k, . . . , 231_p. In addition, each of the groups of amplitude data
from each of the time and frequency conversion processing means 220_0,
220_1, . . . , 220_k, . . . , 220_p is input to each of the modulators
231_0, 231_1, . . . , 231_k, . . . , 231_p. Each of the modulators 231_0,
231_1, . . . , 231_k, . . . , 231_p amplitude modulates each of the cosine
waves from each of the cosine signal generators 230_0, 230_1, . . . ,
230_k, . . . , 230_p with the amplitudes that are expressed by each of the
groups of amplitude data from each of the time and frequency conversion
means 220_0, 220_1, . . . , 220_k, . . . , 220_p. In this manner, the
waveforms in which the temporal changes of the frequency and amplitude of
each band have been reproduced after the temporal change rates have been
adjusted are reproduced. In addition, all of these waveforms that have
been reproduced are combined in the combining section 230. In this manner,
the waveform y(n) in which the original waveform that has been input is
compressed or expanded in the direction of the temporal axis.
Since with the waveform reproduction apparatus of the second preferred
embodiment, the waveform y(n) is obtained by means of the processing as
above, compared to a waveform reproduction apparatus that employs a
cross-fade format with which the waveform data that express the original
waveform are directly jump read or repetitively read out and cross-fade
processed, such things as fluctuation and ripples due to a shift in the
phase that is produced in the vicinity of the discontinuous areas are
reduced. In addition, compared to the format that employs a waveform
reproduction apparatus of the past known as a phase vocoder, the temporal
change rate of the amplitude is adjusted so that it is longer than that
period while maintaining the period of the periodic change of the
amplitude, even in those cases where, in a certain frequency band, the
frequencies that express the multiple number of harmonics that are
adjacent and are integer multiples of the fundamental frequency that
corresponds to the fundamental period exist in multiple numbers, without
compressing or expanding the fundamental period, it is possible to prevent
the breakdown of the harmonic relationships of the original waveform.
Accordingly, it is possible to improve the sound quality of musical tones
etc. that are expressed by waveforms that have been compressed or expanded
in the direction of the temporal axis.
Incidentally, in the first and second preferred embodiments, an explanation
was given with illustrations in which the operator groups comprised
expansion and compression switches and reproduction switches. However,
with the present invention, it is possible to have it comprise a keyboard
and a modulation wheel and apply it also to a system in which the pitch
and time are changed by the direction from the keyboard etc. of analysis
parameters or the audio data of each band that have been analyzed and
which include the production of musical tones in which the generation of
sounds is instantaneous.
Top